id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_qe6lb6u4rfcq3faeborafelgze	Ranko Gacesa	Machine learning can differentiate venom toxins from other proteins having non-toxic physiological functions	2016	20	.pdf	application/pdf	8809	869	53	Keywords Protein sequences, Biological function, Animal venom, Automatic annotation, (2016), Machine learning can differentiate venom toxins from other proteins having non-toxic Moderate and Hard datasets); a sequence is classified as a toxin if the BLAST search ToxClassifier meta-classifier was calibrated by evaluating prediction score versus performance for each animal training set and for summary dataset constructed by combining Table 1 Prediction accuracy on positive and negative datasets, as well as range of measurements calculated for all test data, and described in detail in 'Methods.' Annotation models used as classifier inputs either: the frequency of amino acids (TBSim) or combinations of two amino-acids (BIF); the presence of absence or 'Tox-Bits' Table 2 Performance for selected annotation models and published toxin prediction tools. or BLAST (triBLAST) comparisons with the UniProtKB/SwissProt-ToxProt sequences supplemented with non-toxin sequences from the 'Moderate' and 'Hard' datasets lists comparison of classification performance for calibrated ToxClassifier to BLAST based annotation models and ClanTox toxin prediction server.	./cache/work_qe6lb6u4rfcq3faeborafelgze.pdf	./txt/work_qe6lb6u4rfcq3faeborafelgze.txt